Development of an AI-Based Risk Prediction Model for Sepsis Mortality Using Routine Laboratory Parameters

Abstract

Objective: To develop an AI-based risk prediction model for sepsis mortality using routine laboratory parameters among patients admitted to teaching hospitals of Lahore.

Study Design: Multicenter hospital-based analytical study
Place and Duration of Study: This study was conducted at the Amna Inayat Medical College, Lahore from December 2024 to February 2026.

Methods: This was a multicenter hospital-based analytical study conducted at teaching hospitals of Lahore, including 250 adult patients diagnosed with sepsis according to the Sepsis-3 criteria.

Results: The mean age of patients was 56.8 ± 17.2 years, and overall mortality was 29.6%. Non-survivors had significantly higher serum lactate (5.4 ± 1.9 mmol/L), creatinine (2.4 ± 1.0 mg/dL), procalcitonin (10.4 ± 3.6
ng/mL), and C-reactive protein levels compared with survivors (p <0.001). Gradient boosting demonstrated the best predictive performance with 88.0% accuracy and an AUC of 0.93.

Conclusion: AI-based models using routine laboratory parameters provide accurate prediction of sepsis mortality and may assist clinicians in early identification of high-risk patients.